The present invention pertains to image signal encoding; in particular, it pertains to encoding technology that can be applied to MPEG.
Drawing 1 will be used to explain the encoding method for a conventional MPEG encoder. The coding sequence for the input original images is rearranged in picture-rearranging unit 1, and motion estimation unit 2 estimates the motion of the image over time and outputs a motion vector detected in a prescribed coding mode. Differential calculation unit 3 outputs the difference between the xe2x80x98uncompressed imagexe2x80x99 output by motion estimation unit 2 and the compressed image for which the motion is predicted. DCT unit 4 DCTs that differential output, and quantization unit 5 quantizes that DCT output using a bit rate control value calculated from the usage rate for buffer 10. Variable-length coding unit 6 converts the quantization value to a variable-length code. Inverse quantization unit 7 inverse-quantizes the quantization value using the aforementioned bit rate control value, and it is inverse DCT-ed at IDCT unit 8. Multiplexer unit 9 multiplexes the outputs from variable-length coding unit 6 and motion estimation unit 2, sending the result to buffer 10, and buffer 10 stores that input and outputs that data as a bit stream. Adder unit 11 adds the output from IDCT unit 8 and the output from motion estimation unit 13. Memory 12 stores the compressed image. Motion prediction unit 13 predicts and outputs the optimal motion using the compressed image, the motion vector, and the coding mode.
As explained above, with the conventional MPEG encoding method, the bit rate control value used for quantization is calculated using only the usage rate for buffer 10, so quantization occurs independently of the picture conditions (an image with rapid motion, a still image, etc.), and a visually satisfactory image compression is not possible. For example, as in Drawing 2, when the input image is an image in which there is a fountain in the background, with multiple water motion, and there is a person standing still in the foreground, and when the visual characteristics of the person are considered, since the fountain""s water is a xe2x80x98moving body,xe2x80x99 the blurring of the image is not noticeable, but the stationary person is a still image, so the blurring of the image is noticeable. With the conventional MPEG encoding method, estimation and prediction of the motion for the water portion is difficult, so the differential output from differential calculation unit 3 increases, and a large amount of information is generated by the DCT unit. In this condition, the bit rate is controlled such that the usage rate for buffer 10 is kept constant, so this is controlled by quantization unit 5. In other words, the quantization step is rough, and feedback operates such that a large amount of information is not generated with respect to the output of quantization unit 5. For the buffer usage rate, which is the feedback parameter, an average of 15 images, for example, is taken, so the roughness of the quantization step is without regard to the water portion of the image or the portion of the image with a person""s face, and the person""s face blurs, becoming a visually xe2x80x98unsatisfactory image.xe2x80x99
The MPEG encoder of the present invention is characterized in that the still portions and the moving portions are detected, and the quantization step of the portion being corrected is made smaller. In addition, that decision is made using the image error value and the motion vector value from the motion estimation unit. Furthermore, in that decision a time filter is applied to the still image information. Furthermore, the standard vector value used to detect still image portions is changed according to the time difference for 2 images.